DRS_AI / app.py
dschandra's picture
Update app.py
449d194 verified
raw
history blame
10.8 kB
import cv2
import numpy as np
import torch
from ultralytics import YOLO
import gradio as gr
from scipy.interpolate import interp1d
from scipy.ndimage import uniform_filter1d
import uuid
import os
# Load the trained YOLOv8n model
model = YOLO("best.pt")
# Constants for LBW decision and video processing
STUMPS_WIDTH = 0.2286 # meters (width of stumps)
FRAME_RATE = 20 # Input video frame rate
SLOW_MOTION_FACTOR = 2 # Reduced for faster output
CONF_THRESHOLD = 0.3 # Increased for better detection
PITCH_ZONE_Y = 0.8 # Adjusted for pitch near stumps
IMPACT_ZONE_Y = 0.7 # Adjusted for impact near batsman leg
IMPACT_DELTA_Y = 20 # Reduced for finer impact detection
STUMPS_HEIGHT = 0.711 # meters (height of stumps)
def process_video(video_path):
if not os.path.exists(video_path):
return [], [], [], "Error: Video file not found"
cap = cv2.VideoCapture(video_path)
frames = []
ball_positions = []
detection_frames = []
debug_log = []
frame_count = 0
while cap.isOpened():
ret, frame = cap.read()
if not ret:
break
# Process every frame for better tracking
frames.append(frame.copy())
# Preprocess frame for better detection
frame = cv2.convertScaleAbs(frame, alpha=1.2, beta=10) # Enhance contrast
results = model.predict(frame, conf=CONF_THRESHOLD)
detections = [det for det in results[0].boxes if det.cls == 0]
if len(detections) == 1:
x1, y1, x2, y2 = detections[0].xyxy[0].cpu().numpy()
ball_positions.append([(x1 + x2) / 2, (y1 + y2) / 2])
detection_frames.append(len(frames) - 1)
cv2.rectangle(frame, (int(x1), int(y1)), (int(x2), int(y2)), (0, 255, 0), 2)
frames[-1] = frame
debug_log.append(f"Frame {frame_count}: {len(detections)} ball detections")
frame_count += 1
cap.release()
if not ball_positions:
debug_log.append("No valid single-ball detections in any frame")
else:
debug_log.append(f"Total valid single-ball detections: {len(ball_positions)}")
return frames, ball_positions, detection_frames, "\n".join(debug_log)
def estimate_trajectory(ball_positions, detection_frames, frames):
if len(ball_positions) < 2:
return None, None, None, None, None, None, "Error: Fewer than 2 valid single-ball detections for trajectory"
frame_height = frames[0].shape[0]
# Smooth coordinates with moving average
window_size = 3
x_coords = uniform_filter1d([pos[0] for pos in ball_positions], size=window_size, mode='nearest')
y_coords = uniform_filter1d([pos[1] for pos in ball_positions], size=window_size, mode='nearest')
times = np.array([i / FRAME_RATE for i in range(len(ball_positions))])
pitch_idx = 0
for i, y in enumerate(y_coords):
if y > frame_height * PITCH_ZONE_Y:
pitch_idx = i
break
pitch_point = ball_positions[pitch_idx]
pitch_frame = detection_frames[pitch_idx]
impact_idx = None
for i in range(1, len(y_coords)):
if (y_coords[i] > frame_height * IMPACT_ZONE_Y and
abs(y_coords[i] - y_coords[i-1]) > IMPACT_DELTA_Y):
impact_idx = i
break
if impact_idx is None:
impact_idx = len(y_coords) - 1
impact_point = ball_positions[impact_idx]
impact_frame = detection_frames[impact_idx]
x_coords = x_coords[:impact_idx + 1]
y_coords = y_coords[:impact_idx + 1]
times = times[:impact_idx + 1]
try:
fx = interp1d(times, x_coords, kind='linear', fill_value="extrapolate")
fy = interp1d(times, y_coords, kind='quadratic', fill_value="extrapolate")
except Exception as e:
return None, None, None, None, None, None, f"Error in trajectory interpolation: {str(e)}"
vis_trajectory = list(zip(x_coords, y_coords))
t_full = np.linspace(times[0], times[-1] + 0.5, len(times) + 5)
x_full = fx(t_full)
y_full = fy(t_full)
full_trajectory = list(zip(x_full, y_full))
debug_log = (f"Trajectory estimated successfully\n"
f"Pitch point at frame {pitch_frame + 1}: ({pitch_point[0]:.1f}, {pitch_point[1]:.1f})\n"
f"Impact point at frame {impact_frame + 1}: ({impact_point[0]:.1f}, {impact_point[1]:.1f})")
return full_trajectory, vis_trajectory, pitch_point, pitch_frame, impact_point, impact_frame, debug_log
def lbw_decision(ball_positions, full_trajectory, frames, pitch_point, impact_point):
if not frames:
return "Error: No frames processed", None, None, None
if not full_trajectory or len(ball_positions) < 2:
return "Not enough data (insufficient valid single-ball detections)", None, None, None
frame_height, frame_width = frames[0].shape[:2]
stumps_x = frame_width / 2
stumps_y = frame_height * 0.8 # Adjusted to align with pitch
stumps_width_pixels = frame_width * (STUMPS_WIDTH / 3.0)
batsman_area_y = frame_height * 0.7
pitch_x, pitch_y = pitch_point
impact_x, impact_y = impact_point
in_line_threshold = stumps_width_pixels / 2
if pitch_x < stumps_x - in_line_threshold or pitch_x > stumps_x + in_line_threshold:
return f"Not Out (Pitched outside line at x: {pitch_x:.1f}, y: {pitch_y:.1f})", full_trajectory, pitch_point, impact_point
if impact_y < batsman_area_y or impact_x < stumps_x - in_line_threshold or impact_x > stumps_x + in_line_threshold:
return f"Not Out (Impact outside line or above batsman at x: {impact_x:.1f}, y: {impact_y:.1f})", full_trajectory, pitch_point, impact_point
hit_stumps = False
for x, y in full_trajectory:
if (abs(x - stumps_x) < in_line_threshold and
abs(y - stumps_y) < frame_height * 0.1):
hit_stumps = True
break
if hit_stumps:
if abs(x - stumps_x) < in_line_threshold * 0.1:
return f"Umpire's Call - Not Out (Ball clips stumps, Pitch at x: {pitch_x:.1f}, y: {pitch_y:.1f}, Impact at x: {impact_x:.1f}, y: {impact_y:.1f})", full_trajectory, pitch_point, impact_point
return f"Out (Ball hits stumps, Pitch at x: {pitch_x:.1f}, y: {pitch_y:.1f}, Impact at x: {impact_x:.1f}, y: {impact_y:.1f})", full_trajectory, pitch_point, impact_point
return f"Not Out (Missing stumps, Pitch at x: {pitch_x:.1f}, y: {pitch_y:.1f}, Impact at x: {impact_x:.1f}, y: {impact_y:.1f})", full_trajectory, pitch_point, impact_point
def generate_slow_motion(frames, vis_trajectory, pitch_point, pitch_frame, impact_point, impact_frame, detection_frames, output_path, decision):
if not frames:
return None
frame_height, frame_width = frames[0].shape[:2]
stumps_x = frame_width / 2
stumps_y = frame_height * 0.8 # Align with pitch
stumps_width_pixels = frame_width * (STUMPS_WIDTH / 3.0)
stumps_height_pixels = frame_height * (STUMPS_HEIGHT / 3.0)
fourcc = cv2.VideoWriter_fourcc(*'mp4v')
out = cv2.VideoWriter(output_path, fourcc, FRAME_RATE / SLOW_MOTION_FACTOR, (frame_width, frame_height))
trajectory_points = np.array(vis_trajectory, dtype=np.int32).reshape((-1, 1, 2))
for i, frame in enumerate(frames):
# Draw stumps outline
cv2.line(frame, (int(stumps_x - stumps_width_pixels / 2), int(stumps_y)),
(int(stumps_x + stumps_width_pixels / 2), int(stumps_y)), (255, 255, 255), 2)
cv2.line(frame, (int(stumps_x - stumps_width_pixels / 2), int(stumps_y - stumps_height_pixels)),
(int(stumps_x - stumps_width_pixels / 2), int(stumps_y)), (255, 255, 255), 2)
cv2.line(frame, (int(stumps_x + stumps_width_pixels / 2), int(stumps_y - stumps_height_pixels)),
(int(stumps_x + stumps_width_pixels / 2), int(stumps_y)), (255, 255, 255), 2)
# Draw crease line at stumps
cv2.line(frame, (int(stumps_x - stumps_width_pixels / 2), int(stumps_y)),
(int(stumps_x + stumps_width_pixels / 2), int(stumps_y)), (255, 255, 0), 2)
if i in detection_frames and trajectory_points.size > 0:
idx = detection_frames.index(i) + 1
if idx <= len(trajectory_points):
cv2.polylines(frame, [trajectory_points[:idx]], False, (0, 0, 255), 2) # Blue trajectory
if pitch_point and i == pitch_frame:
x, y = pitch_point
cv2.circle(frame, (int(x), int(y)), 8, (0, 255, 0), -1) # Green for pitching
cv2.putText(frame, "Pitching", (int(x) + 10, int(y) - 10),
cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0), 1)
if impact_point and i == impact_frame:
x, y = impact_point
cv2.circle(frame, (int(x), int(y)), 8, (0, 0, 255), -1) # Red for impact
cv2.putText(frame, "Impact", (int(x) + 10, int(y) + 20),
cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 255), 1)
if impact_point and i == impact_frame and "Out" in decision:
cv2.putText(frame, "Wickets", (int(stumps_x) - 50, int(stumps_y) - 20),
cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 165, 255), 1) # Orange for wickets
for _ in range(SLOW_MOTION_FACTOR):
out.write(frame)
out.release()
return output_path
def drs_review(video):
frames, ball_positions, detection_frames, debug_log = process_video(video)
if not frames:
return f"Error: Failed to process video\nDebug Log:\n{debug_log}", None
full_trajectory, vis_trajectory, pitch_point, pitch_frame, impact_point, impact_frame, trajectory_log = estimate_trajectory(ball_positions, detection_frames, frames)
decision, full_trajectory, pitch_point, impact_point = lbw_decision(ball_positions, full_trajectory, frames, pitch_point, impact_point)
output_path = f"output_{uuid.uuid4()}.mp4"
slow_motion_path = generate_slow_motion(frames, vis_trajectory, pitch_point, pitch_frame, impact_point, impact_frame, detection_frames, output_path, decision)
debug_output = f"{debug_log}\n{trajectory_log}"
return f"DRS Decision: {decision}\nDebug Log:\n{debug_output}", slow_motion_path
# Gradio interface
iface = gr.Interface(
fn=drs_review,
inputs=gr.Video(label="Upload Video Clip"),
outputs=[
gr.Textbox(label="DRS Decision and Debug Log"),
gr.Video(label="Optimized Slow-Motion Replay with Pitching (Green), Impact (Red), Wickets (Orange), Stumps (White), Crease (Yellow)")
],
title="AI-Powered DRS for LBW in Local Cricket",
description="Upload a video clip of a cricket delivery to get an LBW decision and optimized slow-motion replay showing pitching (green circle), impact (red circle), wickets (orange text), stumps (white outline), and crease line (yellow line)."
)
if __name__ == "__main__":
iface.launch()